Cannot pull from Pronunciation audio if Attach was used instead of Audio WS

Pulling the first audiorecording from Pronunciations only works if using an audio input system. (Lexeme/Citation Form with audio work too.) Simply attaching a file to a Pronunciation doesn't work. (There's a problem in the XPath package being used: it has trouble returning XML attributes, which is where that filename is stored in LIFT.)
